How evasion works

https://forums.galaxy-of-heroes.starwars.ea.com/discussion/46998/reality-of-ai-dodge-complete-statistics/p1

I believe that might be close. The way u calculated player.
Personally I have seen base dodge rate much lower than 6.5 percent.
And I don't think 1.065 x 1.15 is technically mathematically correct.

1.15 is 115%, which is 15% more than 100% of base rate, which is fine.

But 1.065 means dodge rate of 106.5% is the base rate, which is obviously not correct.

I believe either
1. Dodge bonus is added to base dodge rate. 6.5% plus 15% =21.5%. To me this is symantically the best mathematical representation of the wording of the bonus.

2. I theorize Dodge rate could also be calculated as a percentage of the HIT rate. I.e. base chance to be hit is 100% minus 6.5% (in your example). 93.5% is the base chance to be hit. Then the dodge bonus would be 0.85 x 0.935 to get chance to be hit, or lets call it HR. 100 (1 - HR) = chance to dodge in %.

But this is just my theory, and there is no real reason to do it this way, other than it is mathematically correct and comes out close, as well.

If you take it ato face value it might even be interpreted to mean
3. 0.065 x 1.15 for a dodge rate of about about 7.5%, which is obviously not correct.

Then again it could be
4. Your way, which is in the right zip code, but as I said is technically not correct, mathematically, at all. But it works out pretty close. And who knows what the coders found easiest and/or if they are math majors.

But I dont see your AI calculation being close in practice. Actually I see no AI advantage, at all. So far. I don't know how u even came up with the hypothesis to begin with. It seems like it is pulled out of thin air.

In the absence of any statistical data, I'm forced to offer my Theory of Dodge. Feel free to review it as you please.

This is a terrible way to come up with a hypothesis, let alone a theory. :)

BTW, lumi and ben proc evasion up. I have noticed two really awesome dodge streaks in my life. One was AI lumi. One was mine. And Evasion up was active during parts of both of those tears. I am sure AI ben has had a lot of really good streaks, too, but I don't pay attention. It's on auto, by then.

I always avoid targeting lumi when she has evasion up. (And Ben is last and on auto, so w/e).

@LastJedi while my parameters are a thumb suck, the maths is in fact "technically mathematically" correct.

In the absence of any statistical data, I'm forced to offer my Theory of Dodge. Feel free to review it as you please.

This is a terrible way to come up with a hypothesis, let alone a theory. :)

You are quite correct; it is in fact a model.

In the stats that @LastJedi and I gathered and posted in his thread the "base" evasion rate seemed to be about 3-4% and evasion leaders seemed to ADD 15% to that to make it roughly 18-20%. There appeared to be very little to NO difference between the player and AI.

I encourage you to join us in recording some actual data on this so we can get a better more accurate understanding and come up with closer models.

"MayCM;488140" wrote:
So you are just making up numbers? I don't get it.

I have attempted to build a predictive model that explains how evasion works. I am not saying the numbers I use are correct, in fact they are purely guesses. The point is that there must be a base evasion rate and that multipliers are applied to this rate in a certain way, to give an actual evasion rate.

I do state that the AI has a higher rate of evasion---this is my perception and may be incorrect. However, @DjangoQuik, any data gathered on an ad-hoc basis necessarily have the status being anecdotal, as it cannot be independently verified. Such attempts, with respect, are then useless.

Data quoted by a reputable source, such as the game publisher, would in my opinion be acceptable

"Scourge194;488173" wrote:

any data gathered on an ad-hoc basis necessarily have the status being anecdotal, as it cannot be independently verified. Such attempts, with respect, are then useless.

Data quoted by a reputable source, such as the game publisher, would in my opinion be acceptable

Hundreds of data points are far from useless. It is a reasonable start at producing empirical evidence to illustrate some aspects of the evasion system, such as how the supposed AI bonus is purely observation bias.

The more data points collected, the more we can isolate and diminish anomalies and get closer to the statistics as intended by the developers. Unless they provide the actual calculations used, this is the best chance we have at identifying the truth about evasion statistics and is far more useful that making up a theory with no evidence, using numbers plucked from thin air and putting it all in a thread entitled "How evasion works".

You wanna figure out how it actually works without the devs help, take some stats down yourself instead of making stuff up because it makes sense to you.

1.065 x 1.15 = 1.22475.

That statement is mathematically correct.

But in the context, it makes no sense.

I have heard other people state like fact that "in mmo/rpg, this is how multipliers are added."

What u have here, mathematically, is 106.5% chance to evade x 115% = 122.5% chance to evade. That is nonsense.

You do not arbitrarily add 100% to the base dodge rate before doing the multiplication, then subtract it, after. The confusion may come from the 115%. In this case you do it because the 15% is an increase from 100% of the baseline rate. You do not add the 100% then mutiply, then subtract it. After multiplying, you are done. It is not arbitrarily added then removed.

Not all mmo rpg experts paid atten5ion in high school. It may indeed be the way Runescape works, or whatnot. But that doesn't make any mathematical sense. In fact, what you are doing here is straight adding +15% and then adding an additional 15% of base on top of that. There is no (competent) mathematician in the world that will tell you this equates to increasing the base rate by 15%, anyway you color it.

The AI totally gets an RNG bonus. Not just evade, but all those RNG chances

Anyone claiming ai does get a bonus - please take note of multiple battles and actually count how often which characters dodge. Doing this will instantly disprove what you are "observing" - multiple people have now done this and found the same result.

"Timitock;489911" wrote:

"DjangoQuik;489907" wrote:
Anyone claiming ai does get a bonus - please take note of multiple battles and actually count how often which characters dodge. Doing this will instantly disprove what you are "observing" - multiple people have now done this and found the same result.

I have. The AI dodges more, rezzes more, assists more, and sticks more debuffs.

This is counter to others' recorded stats. Would be good to see your numbers.